The economic significance of commerce in urban areas and its close association with the livelihoods of urban residents is a well-established fact. Commercial attractiveness, a vital measure of commercial facilities, indicates the potential consumer base and possible financial benefits. The interaction between specific features of the built environment and human activities is complex and subtle. Within the commercial environment, this relationship mainly refers to the capacity of environmental features in areas with commercial facilities to attract urban residents. As a result, it has garnered increasing attention from scholars and businesses alike.

Despite this, research on the relationship between commercial attractiveness and the built environment is still an underexplored topic. Existing studies primarily concentrate on the store level and the impact of the physical environment on customer satisfaction, with only a few examining the impact of the environment on commercial attractiveness on multiple scales, such as streets, blocks, and cities. Furthermore, commercial attractiveness and the related factors vary among different cities and spaces, making generalizations difficult. Unfortunately, a majority of the existing studies are region-specific, resulting in weak pertinence and specificity of research results. Finally, the specific function of commercial facilities makes their performance highly time-sensitive. However, existing studies predominantly focus on the spatial relationship, with little attention paid to the difference in time.

To address these research gaps, this study measured the built environment factors based on multi-source urban data, and systematically identified and analyzed the factors influencing commercial attractiveness at the micro and macro level. First, spatial features within the city limits were selected to formulate the metrics for the built environment, and the patterns of human activity were measured quantitively to compose the proxy of commercial attractiveness. Second, macro features were collected on a block basis, and a machine learning model was used to rank the importance of the features at the macro level to achieve predictions of commercial attractiveness. Then, the associations at the micro level were discussed by using streets as basic analysis units. Spatially, relationships between built environment features and commercial attractiveness were established through the classification of diverse street characteristics in multiple districts. Temporally, such associations were investigated by dividing the time of day in a single district into multiple periods.

At the macro level, the focus is on the correlation between environmental features and the commercial attractiveness of various block units in urban areas. The study objectively measures the environmental features and commercial attractiveness of each neighborhood and uses three machine-learning methods (linear regression, k-nearest neighborhood and random forest) to establish prediction models and evaluate the significance of features. Based on urban form typology, the city blocks are classified according to type, and each type is tested with the prediction model to compare the difference in prediction accuracy.

At the micro level, the study aims to investigate the relationship between the street features of commercial districts and commercial attractiveness in different spatial and temporal distributions. In the spatial dimension, the study collected and quantified the perceptible environmental features of the street and population density data in four different commercial districts. The streets were clustered into five categories according to the feature similarity. The weighted least squares (WLS) model was established in each street cluster to reveal the correlations with commercial attractiveness in diverse street spaces. In the temporal dimension, the street features were collected in a single commercial district to connect with the pedestrian volume representing commercial attractiveness. The generalized linear models (GLM) for four time periods analyzed the distribution of pedestrians and their correlated street features in commercial districts at diverse times.

The study revealed that the relationship between commercial attractiveness and urban environmental features varies significantly across different scales, time, space and regions. The findings indicate that at the macro level, commercial facility density and scale of the blocks contribute more to the commercial attractiveness. Additionally, the random forest model has a better prediction effect among the three machine learning models and is appropriate for common block typology in Chinese cities. At the micro level, the effects of street form and visual perception are apparent rather than functional density, especially on weekdays. Furthermore, the impact of visual perception is more significant in streets around subway stations, and convenient transportation is vital to streets with dense function and open spaces. The functional density has a greater impact on streets when there is better greening and walkability.

This research sheds light on the intricate relationship between the built environment and human activities in commercial settings, taking into account multiple scales and space-time factors. The study highlights the significant role that environmental features play in enhancing the commercial attractiveness of a location. This study contributes to the existing literature on urban environments and commercial districts, presenting crucial insights for future renewal, development, and revitalization of commercial districts. The results of this research can help stakeholders make informed decisions about the design and development of commercial spaces, which can ultimately contribute to the economic growth and well-being of communities.